Live data from Hacker News

Shall We Play a Market Timing Game? (2018)

engaging-data.com

11–20 of 52 posts

Re: Shall We Play a Market Timing Game? (2018)

#11
post #8

Generally, in markets that were part bearish (like the 80s), I made very good returns, between 10 % and 30 % above the index. As a very simple rule, after 2 days of a falling index, sell for exactly 1 day. This works mainly because there are enough consecutive 3 days of a falling index. However, if the market trend is overwhelmingly bullish, that does not work anymore, because there are not enough triplets of falling…

> You cannot be better then the index in rising markets, but you could be better in falling markets.

That's because the only actions you can take in the game are "buy the index" and "sell the index".

As soon as there's more than one price in the market (even, say, one index tracking DJIA and another one tracking S&P 500), it's possible to beat the market average while prices are rising.

Re: Shall We Play a Market Timing Game? (2018)

#12
post #9

This just shows that so-called ‘technical’ analysis with no context is about as useful as trading based on horoscopes. Add in some information like ‘a new pandemic threatens to shut the world economy for months and kill tens of millions of people’ and suddenly this changes.

Technical analysis has validity - it is just patterns and data. It isn't perfect, because outcomes are still variable, and also consist of independent human decisions.

A pandemic is an edge case... However, markets react much more quickly than in the past (algorithms, global data, instant analysis of that data) - that technical analysis short term timelines have compressed.

If you want statistical breakdowns of each pattern, their movements, and outcomes - this book is a great reference. https://www.amazon.com/Encyclopedia-Chart-Patterns-Thomas-Bu...

Re: Shall We Play a Market Timing Game? (2018)

#13
post #8

Generally, in markets that were part bearish (like the 80s), I made very good returns, between 10 % and 30 % above the index. As a very simple rule, after 2 days of a falling index, sell for exactly 1 day. This works mainly because there are enough consecutive 3 days of a falling index. However, if the market trend is overwhelmingly bullish, that does not work anymore, because there are not enough triplets of falling…

> You cannot be better then the index in rising markets, but you could be better in falling markets. That's because the only actions you can take in the game are "buy the index" and "sell the index". As soon as there's more than one price in the market (even, say, one index tracking DJIA and another one tracking S&P 500), it's possible to beat the market average while prices are rising.

Not saying you are wrong, but what algorithm would you use to achieve that?

Re: Shall We Play a Market Timing Game? (2018)

#14
Sure. It's a really bad idea to try to time the market on a day to day basis. There's two sides to every trade, so you're generally playing poker against professionals with armies of quants.

But this game really just proves that you can't time the market while knowing nothing about the outside world.

Headlines sometimes matter.

Not all the time. Talking heads generally overstate how much one random speech matters, "politician X says Y, therefore stocks are reacting" is generally just over-analyzing noise. It's annoying to see post hoc rationalizations the norm in financial... in all news. People claim causation for anything they happened to read.

But occasionally, once a decade, say? Headlines do send a strong signal. "Crisis on Wall Street as Lehman Totters" was a headline that came just before the biggest cliff in 2008.

In February 2020 we didn't know as much as we do now, but people already started talking about difficulties in containment. China--which likes itself some economic growth (if only to keep the Party going)--decided a near total economic shutdown was necessary. The virus was already in a few dozen countries and Singapore was surrounded by container ships that couldn't dock.

So let's split out two separate claims:

1) The weak efficient markets hypothesis: you can't time the market blind.

2) Strong EMH: you can't time the market, not even once in a while, with your eyes wide open.

In defense of (2), the housing crisis really started in October 2007, and ended in February 2009. You can delve for headlines for those moments, but they are way more specious. Here's the real test: what will be the best signal of the rally after the pandemic?

1) China hitting zero active cases, reassuring the world it can be done.

2) Global new case "growth rate" below 1 for two straight weeks? [Growth rate being the ratio of today's new cases to yesterday's new cases, with lower than 1 a tipping point away from exponential growth, and probably the halfway point in the crisis.]

3) China keeping no new local cases even after reopening its economy?

4) Some decision about how to keep airlines solvent?

5) Unemployment nearing historical normals after skyrocketing beyond anything we've ever seen?

Any of these seem plausible. But who knows which will be right? However, if we are attacking strong EMH, we don't really have to time things maximally, we just have to do slightly better than the index. So maybe just wait until the S&P 500 has recovered a quarter of its losses and get in then?

Maybe the general problem with active investing is the financial sector's approach: hire people to study market signals full time and generate algorithms that can trade more and more actively. Full time people are expensive, and tuning algorithms is expensive. So that means you burn through fees (on top of probably not outperforming the market, because you're competing against noise).

If you hire people to do something full time, they will find ways to justify their time. If you hire someone to play rock paper scissors full time, and give them some of the highest bonuses in the world, they will come up with some very nice models. And usually not outperform a random thrower, but give you lots of reports on why and how they'll do better next time.

But if all the daily signals are noise except for one really blaring foghorn once a decade, maybe the better solution would be to hire a part time market hobbyist on a contingency fee. "Hey, if you see a signal that the entire market should be shorted, maybe short the market. You get two trades per decade max. Otherwise, just index and hold."

Probably also wouldn't work, but I really like the idea of some plumber in Poughkeepsie controlling billions of dollars in hedge fund money, you know, just as a side hustle.

Re: Shall We Play a Market Timing Game? (2018)

#16
post #13

Earlier quoted context omitted.

> You cannot be better then the index in rising markets, but you could be better in falling markets. That's because the only actions you can take in the game are "buy the index" and "sell the index". As soon as there's more than one price in the market (even, say, one index tracking DJIA and another one tracking S&P 500), it's possible to beat the market average while prices are rising.

Not saying you are wrong, but what algorithm would you use to achieve that?

Always invest in the one that's increasing faster than the other one.

I'm not saying you can do it without knowledge of the future. I'm responding to the (correct) observation above that it isn't possible to beat the market average while it's rising, no matter what you do, even if you do have knowledge of the future -- as long as the only options you have are "buy the index" and "sell the index".

Re: Shall We Play a Market Timing Game? (2018)

#17
post #15

I predict that the market will see a handful of the largest single day point increases within the next six to ten months. Just a prognostication on my part, it's worth exactly what you paid for it.

Seems legit. We already had one of the largest single day jumps in the stock market the day after one of the largest drops and the day before another one of the largest drops.

Re: Shall We Play a Market Timing Game? (2018)

#18

Sure. It's a really bad idea to try to time the market on a day to day basis. There's two sides to every trade, so you're generally playing poker against professionals with armies of quants. But this game really just proves that you can't time the market while knowing nothing about the outside world. Headlines sometimes matter. Not all the time. Talking heads generally overstate how much one random speech matters, "p…

The Efficient Market Hypothesis is obviously, patently false. The markets cannot agree on the value of an asset from week to week, day, hour, minute or second. Equities in stable businesses with millions of shares traded daily see their prices fluctuate 5%, 10%, 20% intra-day. The tangible value of a company simply does not change that fast. It doesn't.

It's impossible to time the market perfectly every time because that would imply perfect knowledge of the moves of all the participants. By the same token, it's impossible to mis-time the market every time, because then you could just take all the opposite moves and you're back to winning. It is possible to win more than you lose, not by being the smartest, but simply by being smarter than the average participant. Which, thanks to companies like Robinhood putting trading capability into the hands of any naive smartphone owner, has become easier than ever.

Re: Shall We Play a Market Timing Game? (2018)

#19

Sure. It's a really bad idea to try to time the market on a day to day basis. There's two sides to every trade, so you're generally playing poker against professionals with armies of quants. But this game really just proves that you can't time the market while knowing nothing about the outside world. Headlines sometimes matter. Not all the time. Talking heads generally overstate how much one random speech matters, "p…

I believe the strongest, first order, signal for a rally that sticks will be when there's a somewhat effective treatment for the virus that can be deployed in a scalable way. This will mean that the healthcare systems will be under less pressure as they will be operating in a different mode and that will mean that restrictions that have a negative economic impact will be lifted.

Until then, there's always the prospect of resurgence of the virus once restrictions are lifted, and that will keep the lid on the markets.

If there's strong news of effective treatments in the short term, that may kick off a sustained rally soon.

Of course the other things you mention are very valid too and will have positive effects, and may be "the one".

Re: Shall We Play a Market Timing Game? (2018)

#20
> Update: Added a Monte Carlo mode which lets you play with data that is randomly generated from the daily returns of the S&P500. The probability of a daily return being picked is the same probability/frequency that it occurred in the last 68 years.

This mode is rigged.

Any proposal for market timing requires correlated returns. "Technical" traders infer short-term trends form patterns like the shave-and-a-haircut and lovely-lady-humps that are ultimately based on a theory of market psychology, and "fundamental" traders usually make structural observations like price-to-earnings ratios being mean-reverting.

Both of these hypotheses are excluded by construction when market results are constructed by a random draw of daily returns.

Post reply on HN